Atmos. Meas. Tech., 13, 6053–6065, 2020https://doi.org/10.5194/amt-13-6053-2020© Author(s) 2020. This work is distributed underthe Creative Commons Attribution 4.0 License.

Multifactor colorimetric analysis on pH-indicator papers:an optimized approach for direct determination ofambient aerosol pHGuo Li1, Hang Su1, Nan Ma2, Guangjie Zheng1, Uwe Kuhn1, Meng Li1, Thomas Klimach1, Ulrich Pöschl1, andYafang Cheng1

1Max Planck Institute for Chemistry, Mainz, Germany2Institute for Environmental and Climate Research, Jinan University, Guangzhou, China

Correspondence: Hang Su (h.su@mpic.de) and Yafang Cheng (yafang.cheng@mpic.de)

Received: 15 October 2019 – Discussion started: 16 October 2019Revised: 30 August 2020 – Accepted: 4 October 2020 – Published: 13 November 2020

Abstract. Direct measurement of the acidity (pH) of ambi-ent aerosol particles/droplets has long been a challenge foratmospheric scientists. A novel and facile method was in-troduced recently by Craig et al. (2018), where the pH ofsize-resolved aerosol droplets was directly measured by twotypes of pH-indicator papers (pH ranges: 0–2.5 and 2.5–4.5) combined with RGB-based colorimetric analyses usinga model of G−B (G minus B) vs. pH2. Given the widepH range of ambient aerosols, we optimize the RGB-basedcolorimetric analysis on pH papers with a wider detectionrange (pH∼ 0 to 6). Here, we propose a new model to es-tablish the linear relationship between RGB values and pH:pHpredict= a ·Rnormal+ b ·Gnormal+ c ·Bnormal. This modelshows a wider applicability and higher accuracy than thosein previous studies and is thus recommended in future RGB-based colorimetric analyses on pH papers. Moreover, weidentify one type of pH paper (Hydrion® Brilliant pH dipstiks, lot no. 3110, Sigma-Aldrich) that is more applica-ble for ambient aerosols in terms of its wide pH detectionrange (0 to 6) and strong anti-interference capacity. Custom-made impactors are used to collect lab-generated aerosols onthis type of pH paper. Preliminary tests show that, with acollected particle size range of ∼ 0.4–2.2 µm, the pH papermethod can be used to predict aerosol pH with an overall un-certainty ≤ 0.5 units. Based on laboratory tests, a relativelyshort sampling time (∼ 1 to 4 h) is speculated for pH pre-diction of ambient aerosols. More importantly, our design ofthe impactors minimizes potential influences of changed en-vironmental conditions during pH paper photographing pro-

cesses on the predicted aerosol pH. We further show that theroutinely adopted way of using pH color charts to predictaerosol pH may be biased by the mismatch between the stan-dard colors on the color charts and the real colors of investi-gated samples. Thus, instead of using the producer-providedcolor charts, we suggest an in situ calibration of pH paperswith standard pH buffers.

1 Introduction

Aerosol particles have vital impacts on atmospheric chem-istry, human health, and global climate (Pöschl, 2005; Bal-tensperger et al., 2008; Pósfai and Buseck, 2010; von Schnei-demesser et al., 2015; Shiraiwa et al., 2017). Understandingthe basic physicochemical properties of aerosols can provideinsights into various aerosol processes in the atmosphere andmay further help to establish measures against air pollution(Su et al., 2020; Tao et al., 2020). Aerosol acidity, usuallyquantified by aerosol pH, is one of the most important basicproperties of liquid-phase aerosols. Aerosol pH has multipleeffects on the other properties of aerosols, e.g., aerosol com-position (Cheng et al., 2016), reactivity (Gao et al., 2004;Iinuma et al., 2004; Northcross and Jang, 2007), toxicity(Fang et al., 2017; Chowdhury et al., 2018), phase transi-tion (Dallemagne et al., 2016; Losey et al., 2018), and theirrelated climatic effects (Dinar et al., 2008; Hinrichs et al.,2016; Cai et al., 2018). It also plays a critical role during sec-ondary organic aerosol (SOA) formation (e.g., Surratt et al.,

Published by Copernicus Publications on behalf of the European Geosciences Union.

6054 G. Li et al.: Determination of aerosol pH by multifactor colorimetric analysis

2007; Gaston et al., 2014; Han et al., 2016) and in many otherchemical processes in the atmosphere (Hennigan et al., 2015;Cheng et al., 2016; Wang et al., 2016a; Keene et al., 2004;Ahrens et al., 2012).

Despite its essential importance, currently there are fewaerosol pH measurement datasets available. One main rea-son is the small sizes (with an aerodynamic diameter rangeof 2 nm–10 µm; see McNeill, 2017) of these atmospheric par-ticles, rendering measurements of aerosol pH not as easyas for bulk solutions. Moreover, the non-conservative na-ture of H+, i.e., H+ concentrations not scaling in propor-tion to the dilution levels due to buffering effects (Zhenget al., 2020) and the partial dissociation of weak acids, fur-ther makes probing aerosol pH a challenging topic (Henni-gan et al., 2015). For direct measurements of aerosol pH,two types of methods have been employed: filter-based sam-ple extraction (Koutrakis et al., 1988; Keene et al., 2002;Jang et al., 2008) and spectroscopic or microscopic analy-sis (Li and Jang, 2012; Dallemagne et al., 2016; Rindelaubet al., 2016; Craig et al., 2017; Wei et al., 2018). As the for-mer method is offline, it suffers from both poor time resolu-tion and intensive labor (Hennigan et al., 2015). Moreover,it cannot account for the water in the aerosol droplets andinvolves extraction with solvents that can shift the equilib-ria of present ions, leading to high uncertainties. The lattermethod is normally used for laboratory-generated particleswith simple compositions that cannot fully represent ambientaerosols (Craig et al., 2018). Due to these limitations of di-rect measurements, thermodynamic equilibrium models suchas ISORROPIA-II (Fountoukis and Nenes, 2007) and E-AIM(Clegg and Seinfeld, 2006b, a) have been widely used to es-timate the acidity of ambient aerosol droplets, although com-prehensive evaluations of the acidity are hampered by a lackof observational data. Thus, developing new methods to di-rectly measure ambient aerosol pH is imminently needed toconstrain the output of thermodynamic models.

In a recent study, Craig et al. (2018) reported an intrigu-ing way to directly measure aerosol pH using pH-indicatorpapers, which in the past have been the most common andconvenient tool to test the pH of bulk solutions. To mea-sure aerosol pH, the generated size-resolved aqueous aerosolsamples ((NH4)2SO4−H2SO4) were first collected on pH-indicator papers. Then the color of the samples on pH pa-pers was analyzed quantitatively through a colorimetric im-age processing program (Matlab). In this way, the standardpH color chart of the indicator papers was used as a refer-ence to finally derive the aerosol pH. In terms of aerosol sam-pling, Craig et al. (2018) collected aerosols generated in thelaboratory and from ambient air onto pH papers using a mi-croanalysis particle sampler (MPS-3). The MPS-3 had threestages with aerodynamic diameter cutoff sizes (d50) of 2.5–5.0, 0.4–2.5, and< 0.4 µm for stages 1, 2, and 3, respectively,enabling the analysis of size-resolved aerosol pH. An inter-esting finding from their measurements based on both pHpapers and Raman spectroscopy was that, for systems with

pH< 2, the smaller particles (i.e.,< 0.4 and 0.4–2.5 µm) dis-played a markedly lower pH than the larger particles (i.e.,2.5–5 µm). These results were attributed to ammonia parti-tioning and water loss caused by the increased surface-area-to-volume ratios of smaller particles (Craig et al., 2018). Theuse of pH-indicator papers and the related color processingtechnique introduced by Craig et al. (2018) tactfully circum-vents the challenges and difficulties in aerosol pH measure-ments. However, Craig et al. (2018) only reported two typesof pH papers with relatively high precision for pH measure-ments (one with pH range 0–2.5 and the other 2.5–4.5; seeCraig et al., 2018), whereas in the atmosphere the aerosol pHmay vary in a wide range. Note that the authors did indeedemploy another type of pH paper with a larger pH range from0 to 6 for ambient aerosol sampling; unfortunately they foundthat this paper was not compatible with their Matlab script formore quantitative analysis (Craig et al., 2018). Additionally,due to the small area and various shape of different types ofpH papers, the collection of aerosols on these materials isquite distinct from that on commonly used filters. The col-lected particles may induce a color change only on a smallspot (Craig et al., 2018), differing from the color variationon a much larger scale caused by bulk solutions. Moreover,the environment under which aerosols are collected can indi-rectly affect the measured aerosol pH: in an environment dif-ferent from that which the aerosols were originally in, evapo-ration or condensation of water on pH papers might happen,which may further lead to changes in ion activities and/orwater dispersion or homogeneity on pH papers. Thus, to haveaccurate aerosol pH measurements, special techniques or in-struments need to be developed for effective aerosol collec-tion and pH paper color recognition, and meanwhile carefuldesign should be made to avoid potential impacts of variedenvironmental factors on the predicted aerosol pH.

The colorimetric method used by Craig et al. (2018) wasbased on analyzing the red (R), green (G), and blue (B) chan-nels of the sample images, where a linear dependence of thedifference betweenG and B (G−B) on pH2 was found. Ac-cording to trichromatic theory, RGB are the three primarycolors, and their combination in varying proportions can gen-erate any other specific color (Su et al., 2008). The standardRGB scale is represented by the values of R, G, and B, andeach has a range from 0 to 255. For example, the number[0, 0, 0], i.e., R= 0, G= 0, B = 0, corresponds to absoluteblack and [255, 255, 255] to true white. RGB-based imageanalysis has been applied in the fields of inorganic and an-alytical chemistry. For instance, Selva Kumar et al. (2018)found a good linearity between concentrations of thoriumions (Th4+) and the ratio of R and G (R/G), and Wan et al.(2017) reported a relation between bovine serum albumin(BSA) concentrations and the normalized values of R, G,and B, respectively. In these previous studies, different RGBmodels (i.e., ways to interpret the RGB values) were adopted,however with few detailed explanations on the intrinsic rea-sons. To further enhance the reliability and comparability of

Atmos. Meas. Tech., 13, 6053–6065, 2020 https://doi.org/10.5194/amt-13-6053-2020

G. Li et al.: Determination of aerosol pH by multifactor colorimetric analysis 6055

the data associated with RGB analysis, a unified model ormethod to deal with the RGB information is needed, espe-cially for the pH determination of aerosols where a high un-certainty of measured pH values can have a huge impact onthe pH-dependent multiphase chemical processes.

Considering that the pH values of ambient aerosols cancover a wide range (up to ∼ 6) (von Glasow and Sander,2001; Pszenny et al., 2004; Song et al., 2018; Shi et al.,2019), the goal of the present study is to optimize the RGB-based colorimetric analysis on pH-indicator papers for directdetermination of ambient aerosol pH in a wider detectionrange (pH∼ 0 to 6). We thus propose a new way to analyzethe RGB values and establish the relationship between RGBand pH. We further compare our proposed RGB model withthe models used in previous studies in terms of evaluating theestablished linear relationship between RGB and pH. In ad-dition, the routine way of using a pH color chart to derive thepH of samples is inspected, and the results reveal some defi-ciencies of this method. Therefore, we suggest an optimizedway to use pH papers for aerosol pH prediction with higherprecision and accuracy. Nine types of pH papers are testedfor their potential of probing pH of ambient aerosols. Amongthese pH papers, only one type is found to be the most suit-able and is further tested for its capability of predicting thepH of lab-generated aerosols by using two custom-made im-pactors.

2 Materials and methods

2.1 pH-indicator papers

Nine types of pH-indicator papers were adopted in this study.Each type has a pH color chart that is accompanied with thepH papers and is supposed to serve as a reference to quan-tify the pH of a sample through colorimetric analysis. Detailsabout the pH paper detection ranges and the correspondingtype classification used in this work can be found in Table S1in the Supplement. The first two types are the same as usedby Craig et al. (2018), aiming to compare our results withthose from Craig et al. (2018) and validate our colorimetricimage processing method. The others have larger pH detec-tion ranges covering the generally observed pH range of am-bient aerosols. Note that in this study, we mainly focused onthe first five types of pH papers, and the remaining four typeswere also evaluated and compared with the first five types interms of their resistance to chemical interference and poten-tial capability to measure the pH of ambient aerosols.

2.2 pH buffers, aerosol sample solutions, andlab-generated aerosols

To examine the correlation between RGB and pH, eight stan-dard pH buffer solutions were used as purchased; mean-while several other buffers (with different pH values to thepurchased ones) were obtained by mixing the commercial

buffers with solutions of sodium hydroxide or hydrochloricacid (prepared using de-ionized water, 18.2 M�cm). pH val-ues of all the buffers were further checked by a pH bench me-ter (model: HI 2020-02, Hanna Instruments Inc., USA). Priorto the check, the pH meter was calibrated with a three-pointcalibration mode using the standard buffer solutions providedby Hanna Instruments Inc., USA. The measured pH valuesand their standard derivations are listed in Table S2, and themeasured pH values show a small deviation from those spec-ified on the buffer solution bottles, within the displayed un-certainties concomitant with these specified values.

Considering that some inorganic or organic componentsof ambient aerosols might interfere with the dyes on pH pa-pers and cause biased estimation of pH, salt systems withvarying inorganic and/or organic acids common in aerosolsand pH levels (as measured by the pH bench meter) were em-ployed to test the applicability of different types of pH paperscombined with our RGB model. Details about the composi-tion of the tested salt systems can be found in Table S3. Ingeneral, the inorganic systems were similar to those used byCraig et al. (2018). Here, we further tested the influence oforganic acids on pH paper performance by adding organicacids into the inorganic systems. As oxalic acid (C2H2O4)and malonic acid (C3H4O4) were frequently detected in tro-pospheric aerosols and found to be the dominant short di-carboxylic acids in aerosol composition (Abbatt et al., 2005;Falkovich et al., 2005), they were adopted in this study. Forthe solution preparation of each system, varying amounts of1 M inorganic or organic acids were added into 30 mM in-organic salt solution to achieve different pH levels (Surrattet al., 2008; Craig et al., 2018). To prepare the inorganicand organic mixtures, the amount of added organic acidswas generally 2 times larger than the inorganic acids, andthe final salt concentration could be as low as 15 mM due tothe dilution effect of added acids. To prepare the solutions,all chemicals were used as purchased: NaOH (≥ 99.0 %,Roth, Germany), Na2SO4 (≥ 99.0 %, Merck, Germany),NaNO3 (≥ 99.0 %, Merck, Germany), Na2CO3 (≥ 99.5 %,Sigma-Aldrich, USA), (NH4)2SO4 (≥ 99 %, Sigma-Aldrich,USA), NH4NO3 (≥ 98.0 %, Fisher Chemical, USA), MgSO4(> 98 %, neoFroxx GmbH, Germany), H2SO4 (98 %, Merck,Germany), HNO3 (65 %, Merck, Germany), HCl (37 %,Merck, Germany), C2H2O4 · 2H2O (≥ 99 %, Sigma-Aldrich,USA), and C3H4O4 (99 %, Sigma-Aldrich, USA).

To test the feasibility of the colorimetric analysis methodtowards real aerosols, the prepared aerosol sample solutions(i.e., the inorganic and organic mixtures) were further used togenerate aerosol particles through an aerosol generator un-der laboratory conditions. The lab-generated aerosols werecollected onto the type V pH paper through two custom-made impactors, which had different cutoff sizes and wereconnected in series. Before collection, the nebulized aerosolswere first mixed with humidified and HEPA-filtered air toreach a relative humidity (RH) of 90 %± 1.5 % and a totalflow rate of 28.6 Lmin−1. To minimize water exchange be-

https://doi.org/10.5194/amt-13-6053-2020 Atmos. Meas. Tech., 13, 6053–6065, 2020

6056 G. Li et al.: Determination of aerosol pH by multifactor colorimetric analysis

Figure 1. Schematic of using the RGB-based colorimetric method for pH estimation. For the color-chart-calibration method, both the colorchart and the standard buffer samples are imaged into one digital photo for subsequent processing. For the standard-buffer-calibration method,only the standard buffer samples are used for imaging. Note that when using the standard-buffer-calibration results to predict the pH of aerosolsamples, the photographing conditions for the samples are the same as those of the buffer calibration.

tween the generated aerosol flow and the humidified aerosol-free air flow, the RH of the air flow was maintained similarto that of the aerosol flow. With the sampling flow rate of28.6 Lmin−1, the upstream impactor had a cutoff diameter(d50) of ∼ 2.2 µm (identified by a UV-APS, model 3314, TSIInc.) and the downstream impactor had a d50 of ∼ 0.40 µm(identified by an SMPS, model 3082, TSI Inc.). These twoimpactors produced a total pressure drop of 57 mbar in theaerosol line (measured by a digital pressure meter, modelGMH 3111, GHM Messtechnik GmbH, Germany). To val-idate our method, one Wi-Fi endoscope camera was installedon the top of the downstream impactor (with a collected par-ticle size range of 0.40–2.2 µm) to capture the images of onepH paper (5 mm× 5 mm) fixed on the impactor bottom plate.In practice, we could install a camera for each impactor. Inorder to apply our RGB model (Sect. 2.4), a series of stan-dard buffers were also adopted to generate aerosols with thesame experimental configuration mentioned above.

Given that in real ambient cases some light-absorbingparticles, such as black carbon (BC), may interfere withthe displayed color of pH papers and therefore cause bi-ased pH prediction, commercial soot samples (fullerene soot,lot no. L20W054, Alfa Aesar, Germany) were additionallymixed into the aerosol sample solutions for aerosol gen-eration to check their potential impact on the predictedaerosol pH. To achieve this, pure BC suspension was firstprepared with de-ionized water, and then a 15 min ultra-sonic treatment was performed to enhance the dispersionof BC particles inside the suspension. The mass concentra-tion of BC particles (measured under dry conditions with a

RH= 14 %) generated from this suspension was quantifiedby the SMPS as ∼ 240 µgm−3 using the density of fullerenesoot of 1.72 gcm−3 (Kondo et al., 2011). A total of 5 mL ofthis suspension was additionally mixed into 10 mL of pre-prepared aerosol sample solution, and this mixture was fi-nally used for aerosol generation. A total mass concentra-tion of the generated aerosols (measured under dry condi-tions with a RH= 14 %) was determined by the SMPS as∼ 800 µgm−3 using a density of 1.7 gcm−3. This density wasobtained by averaging the densities of different componentsweighted by their respective volume in the aerosol samplesolution mixed with BC. Note that the BC mass fraction was∼ 10 %, representing a typical BC contribution in ambientaerosols (Wang et al., 2016b; Chen et al., 2020).

2.3 Correlation between RGB and pH

Figure 1 shows the procedure of how to use a colorimet-ric analysis to obtain the correlation between RGB and pH.First, 2 µL of liquid samples was dripped onto each piece ofpH paper held by a clean transparent glass plate (with theother side coated by a piece of graph paper). This adoptedsmall volume (2 µL) was based on a general estimation of theavailable amounts of liquid aerosols for aerosol sampling un-der typically polluted conditions (with PM2.5 mass concen-tration around 100 µgm−3) with high RH (60 %–80 %), andassuming an aerosol collection efficiency of 50 % and a sam-pling flow rate of several hundred liters per minute (e.g., canbe achieved by a Tisch Environmental PM2.5 high-volume airsampler; see https://tisch-env.com/high-volume-air-sampler/

Atmos. Meas. Tech., 13, 6053–6065, 2020 https://doi.org/10.5194/amt-13-6053-2020

G. Li et al.: Determination of aerosol pH by multifactor colorimetric analysis 6057

pm2.5, last access: 15 September 2019) with a sampling timeof a few (2–4) hours. Here, the used PM2.5 mass concentra-tion and RH refer to the conditions during haze events whichare frequently occurring in China. For example, during themost severe haze episodes in January 2013, monthly aver-aged PM2.5 concentration in Beijing reached 121 µgm−3 andthe RH was constantly at a level of 60 %–80 % (Zheng et al.,2015). Even though the air quality in China has significantlyimproved in recent years, the number of days with moderatehaze (with a daily mean PM2.5 concentration in the range of100–200 µgm−3) in the North China Plain shows no obvi-ously decreasing trend from 2004 to 2018 with an averageof 113 d (Zhang et al., 2020). Note that, we further estimatedthe minimum sample volume and mass needed to generatea measurable color change on the suggested pH paper. Therelated results are shown below. Then an image of the sam-ple was captured by a smartphone camera (Apple iPhone 5sin this study) immediately. Similar to Craig et al. (2018), thecorresponding color chart of the pH paper was included ineach image to correct for potential influences of variationsin light source and angle during photographing. The digi-tal images were processed by Adobe Photoshop software tocrop a square with 100 pixels× 100 pixels at the center ofthe sample (as well as each color chip on the color chart).The RGB information of the cropped square was then ob-tained and further analyzed by Matlab (The MathWorks, Inc.version R2018b).

2.4 RGB model

Considering that a color is represented by combination ofR, G, and B values, a linear combination of these three pri-mary colors should be able to reflect the characteristics of thecolor and therefore represent the pH related to the color. Suet al. (2008) reported a good correlation between the linearlycombined RGB and the contents of chlorophyll a and lipidin microalgae. To further account for the effect of changinglight intensity on the obtained RGB values, each color chan-nel should be normalized at first (Yadav et al., 2010). Thenormalization can be achieved through Eqs. (1)–(3) shownbelow:

Rnormal = R/(R+G+B), (1)Gnormal =G/(R+G+B), (2)Bnormal = B/(R+G+B), (3)

where R, G, and B are the mean value of each primary coloron the entire 100×100 pixels image. Note that every pixel hasan RGB value vector: [R,G, B ]. Then a model describing thelinear combination of RGB can be given as follows:

pHpredict = aRnormal+ bGnormal+ cBnormal, (4)

where the linear combination aRnormal+bGnormal+cBnormalessentially represents the color information and here can betreated as equivalent to the predicted pH (pHpredict) based on

RGB analysis; a, b and c are the coefficients, which can bedetermined by linear regression analysis through Matlab. Thelinear regression function is expressed as

Y = aX1+ bX2+ cX3, (5)

where Y is the dependent variable vector; X1, X2, and X3 areindependent variable vectors. These vectors can be achievedfrom a standard color chart or a series of buffer samples (withknown pH values) on pH papers: Y is the series of pH values(i.e., reference pH, pHreference) shown on the color chart orof buffer samples (as shown in Fig. 1, the pH papers withdifferent pH buffer solutions are collected together to forma pH series); X1, X2, and X3 are the normalized average ofR, G, and B, respectively, based on analysis on the detectedcolors. As a color chart is normally used as a reference forpH measurements using pH papers, a linear regression anal-ysis on the color chart can provide the coefficient vector [a,b, c] as an answer. Then the same set of coefficient vector(i.e., [a, b, c]) is used to predict the pH (i.e., pHpredict) ofsamples using Eq. (4). Thus, with this RGB model, a lin-ear relationship between RGB-predicted pH (pHpredict) andreference pH (pHreference) is expected for the calibration (asshown in Fig. 1), in order to finally predict the sample pH ona pH paper.

3 Results

3.1 Validation of the new RGB model

As the RGB model (i.e., G−B vs. pH2) used by Craig et al.(2018) produced good linear correlations for the two typesof pH papers that the authors adopted, we first examined thevalidity of this RGB model against the first five types of pHpapers used in this work. Note that here the first two typesof pH papers are the ones used and recommended by Craiget al. (2018). Figure S1 in the Supplement shows the relation-ship between average G−B and pH2 derived from the colorcharts of these five types of pH papers. Relatively good lin-ear correlations can be found for the first three types, which isconsistent with Craig et al. (2018). However, non-monotoniccorrelations are encountered for the last two types of pH pa-pers, which are the ones with wider pH detection ranges (0.5–5.5 and 0–6). These results indicate a limited feasibility of theRGB model proposed by Craig et al. (2018).

Thus, the validity of our new RGB model was furtherchecked through the five types of pH papers. The colorson the color chart for each type of pH paper were first an-alyzed through our RGB model, and then the calculatedpHpredict was compared with the reference pH shown on thecolor chart. As shown in Fig. 2a–e (the “color chart” columnon the left-hand side), good linearity between pHpredict andpHreference can be observed for all these pH paper types.

As mentioned before, besides the RGB model used byCraig et al. (2018), other models have also been adopted to

https://doi.org/10.5194/amt-13-6053-2020 Atmos. Meas. Tech., 13, 6053–6065, 2020

6058 G. Li et al.: Determination of aerosol pH by multifactor colorimetric analysis

Figure 2. Predicted pH (pHpredict) using our RGB model vs. thereference pH shown on the color chart and the pH-meter-probed-pH of the buffer samples (all denoted as pHreference) for the fivedifferent pH papers: (a, f) type I – 0–2.5; (b, g) type II – 2.5–4.5;(c, h) type III – 4.0–7.0; (d, i) type IV – 0.5–5.5; and (e, j) type V –0–6.0. Blue symbols denote the established relationship based oncolor charts only. Red symbols represent the results for 2 µL ofbuffer droplets on pH papers. Both vertical and horizontal error barsrepresent the SD of five to six replicate experiments. Note that theerror bars in most of the panels are smaller than the symbols.

create a linear correlation between RGB and concentrationsof the chemicals of interest in previous colorimetric analyses(Su et al., 2008; Yadav et al., 2010; Wan et al., 2017; SelvaKumar et al., 2018). However, few comparisons have beenmade regarding the goodness of the established linearity us-ing these RGB models. Here we further compared our modelwith the other two models (i.e., R/G vs. pH and G−B vs.pH2) proposed by Selva Kumar et al. (2018) and Craig et al.(2018), respectively, in terms of evaluating their correlationcoefficient R2. Figure S2 (the color chart panel on the left-hand side) displays the R2 of the established linear correla-tion between pHpredict and pHreference when the three modelsare used for the five types of pH papers. For the color-chart-derived linear correlation, the modelG−B vs. pH2 presentspoor goodness of fit for type IV and V pH papers (consis-tent with the results shown in Fig. S1). The model R/G vs.pH shows relatively high R2 for all the pH paper types. Nev-ertheless, this RGB model still underperforms compared toour model. Overall, our RGB model could provide a high R2

(> 0.95) for all the five types of pH papers, which demon-strates the universal validity of our RGB model.

3.2 Calibration with standard buffer solutions

A good linearity, however, may not always be obtained fromthe color chart of some types of pH papers in some pHranges. For example, in the color chart column of Fig. 2,the pHpredict present small but discernable deviations frompHreference for types I, III, and IV pH papers. And the type VpH paper shows even larger differences at both ends of thepH range. A similar phenomenon was also observed in thestudy of Craig et al. (2018) with their RGB model, wherethey argued that the pH paper dye became less effective atthe limits of the pH paper range, due to the pKa values of thedye normally being in the middle of the pH range. But thismay also originate from some color bias due to the differ-ences between the experiment conditions and the ones underwhich the color chart is made by the producer.

Thus, following the same procedure as for the color chart(see Fig. 1), pH papers with samples of a series of 2 µLstandard buffer droplets were also measured. The pH val-ues of the standard buffers were known beforehand and fur-ther checked with a pH meter (also denoted as “pHreference”;see Table S2). Figure 2f–j (the “2 µL buffer” column on theright-hand side) show the comparison between pHpredict andpHreference for the samples of 2 µL buffers. Much better lin-earity between pHpredict and pHreference can be observed forall the five types of pH papers. In particular, the significantdeviation of pHpredict from pHreference found in Fig. 2a–e (thecolor chart column) disappears for the type I and V pH pa-pers. This means that the deviations at the edge of the pHrange in the color-chart-derived calibration curves are mainlydue to the color bias in the color chart itself or caused duringphotographing.

Atmos. Meas. Tech., 13, 6053–6065, 2020 https://doi.org/10.5194/amt-13-6053-2020

G. Li et al.: Determination of aerosol pH by multifactor colorimetric analysis 6059

Actually, even small deviations found in the color-chart-derived calibration curves (the color chart column in Fig. 2)may lead to significant or non-negligible errors in measur-ing aerosol pH. We conducted a case study using the type IVpH paper combined with our RGB model to predict the pHof buffer samples by using the color-chart-derived coeffi-cient vector [a, b, c], i.e., the color-chart-calibration method(Fig. 2d). The blue symbols in Fig. S3 represent pHpredictvs. pHreference of the standard buffer samples. Systemati-cal underestimation of pHpredict can be found at the lowerpHreference values (i.e., pHreference= 1, 1.68 and 2), but aslight overestimation is observed at pHreference= 5. This sig-nificant discrepancy may be attributed to the mismatch be-tween the reference colors on the color chart and the realcolors of the samples, due to the differences between our ex-periment conditions and the ones under which the color chartis made by the producer. This gives us a hint that the coef-ficient vector derived from the color chart is not suitable forpredicting the pH of aerosol samples.

For the established linear relationship using 2 µL standardbuffers, the performances of different RGB models were fur-ther compared and the results are shown in Fig. S2 (the “2 µLbuffer” panel on the right-hand side). Our RGB model stilloutperforms the other two models for all the five types of pHpapers employed in this work. Overall, the good agreementbetween pHpredict and pHreference for all these tested pH pa-pers verifies the wide applicability of our RGB model to thepH paper calibration using standard buffers. In the followingsection, we will examine the quality of predicting sample pHwith the standard-buffer-calibration method.

3.3 pH estimation for aerosol surrogates((NH4)2SO4−H2SO4) with the type IV and V pHpapers

In order to test the feasibility of pH papers with larger pHdetection ranges for pH prediction of aerosols, we furtherused the type IV and V pH papers to estimate the pH of lab-prepared aerosol surrogates ((NH4)2SO4−H2SO4). To min-imize the effect of varying photographing conditions (e.g.,angle, light variation) on the colors of the captured image,experiments were carried out in a cupboard with a con-stant light source. In addition, the pH paper samples as wellas the smartphone were fixed on a small glass plate and ametal holder, respectively, to keep their position unchangedthroughout the experiment. Note that applying the standard-buffer-derived coefficients (Fig. 2i and j) for pH prediction ofsamples required the same constant light source conditionsfor sample imaging processes as for standard buffers.

pHpredict vs. pHreference for the 2 µL droplet samples onthe type IV pH paper are shown in Fig. S4. Generally, thepHpredict by the type IV pH paper is comparable with thepHreference at a lower pH range (i.e., pHreference= 0.46, 1.52and 3.0). However, an anomalous point (highlighted by thearrow in Fig. S4) with 1.5 units of overestimation in pHpredict

can be found at pHreference around 4. This overestimation wasproved to be reproducible by our six replicate experimentsand it was even found for samples of diluted H2SO4 solu-tions with reference pH of around 4 on the type IV pH pa-per. Such overestimation may be due to the chemical inter-ferences caused by the samples or the low buffering levels ofthe samples. Thus, the type IV pH paper is not recommendedfor future pH measurements of aerosols. However, it may stillwork well for the other sample types, such as those found forour self-prepared phosphate buffers (Fig. S4). On the otherhand, the type V pH paper shows decent agreements betweenpHpredict and pHreference within the examined pH range, asshown in Fig. 3a. Moreover, the pHpredict are also comparedwith the results by Craig et al. (2018). The orange and bluebars in Fig. 3a represent the measured pH ranges for the lab-generated (NH4)2SO4−H2SO4 aerosols with particle sizeslarger than 2.5 µm using pH papers (the same as the type Iand II pH papers used here) and Raman spectroscopy, re-spectively.

4 Discussion

4.1 Chemical interference

As mentioned before, aerosol samples with different compo-sitions may have interferences on the indicating color of apH paper and thereby cause its poor performance for aerosolpH prediction, e.g., the overestimation of pH of aerosol sur-rogates ((NH4)2SO4−H2SO4) with the type IV pH paper. Totest the capability of chemical resistance of the type V pH pa-per, we further tested its performance of predicting the pH oflab-prepared aerosol surrogates with varying inorganic or or-ganic compositions that commonly exist in ambient aerosols.

Figure 4 displays pHpredict vs. pHreference for our lab-prepared droplet samples (2 µL) under different pH levels us-ing the type V pH paper. As shown in Fig. 4a, systematicdivergences between pHpredict and pHreference (i.e., overesti-mation of pHpredict when pHreference is in the range of 2.5–3.5 but underestimation of pHpredict when pHreference&4.5)can be found for these tested inorganic systems. Interestinglythese mismatches disappear when the organic acids are in-troduced into these inorganic systems (Fig. 4b), and also forthe cases when the inorganic acids are replaced by organicacids (Fig. S5). In Fig.4b, this good agreement for pHpredictvs. pHreference is observed not only for systems containingoxalic acid (C2H2O4, solid markers) but also for those hav-ing malonic acid (C3H4O4, hollow markers) with an averagedeviation (of pHpredict from pHreference)< 0.5 units. The factthat the existence of organic acids significantly improves thequality of pHpredict may be attributed to the enhanced buffer-ing levels of the samples (Fillion et al., 1999; Li et al., 2016).Actually, good agreement between pHpredict and pHreferenceis found for both the inorganic and organic phosphate sys-tems (Fig. S6) based on our further tests, which is proba-

https://doi.org/10.5194/amt-13-6053-2020 Atmos. Meas. Tech., 13, 6053–6065, 2020

6060 G. Li et al.: Determination of aerosol pH by multifactor colorimetric analysis

Figure 3. pH estimation using the type V pH paper for samples withdifferent volumes: (a) 2 µL and (b) 0.1 µL. pHpredict is calculatedwith the averaged coefficient vector [a, b, c] derived from three tosix replicate experiments with the same amounts of standard buffersas the samples under constant photographing conditions. The errorbars represent the SD of three to six replicate experiments. In (a),the heights of the orange and blue bars indicate the reported pHranges measured with pH papers and Raman spectroscopy, respec-tively, for (NH4)2SO4−H2SO4 aerosols with particle sizes largerthan 2.5 µm in Craig et al. (2018). Each orange or blue bar has thesame pHreference as the red symbol close to it. In (b), for processingthe digital images of the 0.1 µL samples, a square with 20×20 pixelsat the center of the samples is cropped for subsequent colorimetricanalyses.

bly due to the high buffering levels of these systems main-tained by the phosphate itself (Hourant, 2004). Nevertheless,the solvent effect of the added organics on acid dissociationequilibria may also play a role (Padró et al., 2012). The de-tailed mechanisms may need to be explored in future stud-ies. Given the large contribution of organics (Jimenez et al.,2009) and the well-known dominance of both organic acids(i.e., oxalic acid and malonic acid) in ambient aerosols (Ab-batt et al., 2005; Falkovich et al., 2005), the potential inter-ferences found for the inorganic systems can be expected tovanish when organics are concomitant under ambient con-ditions. Additionally, the interference check was also per-

Figure 4. pH estimation using the type V pH paper for salt sys-tems with only inorganic acids (a) and both inorganic and organicacids (b). pHpredict is calculated with the averaged coefficient vector[a, b, c] derived from three replicate calibration experiments withstandard buffers and under constant photographing conditions. Theerror bars represent the SD of three to four replicate experiments.The dotted line in (a) is used to guide the eye.

formed for the other pH paper types (type III and VI–IX)that have larger pH detection ranges. Similar to the type IVpH paper, significant deviations of pHpredict from pHreference(≥ 1.5 units) were observed for these types (see the Supple-ment and Fig. S7).

4.2 BC interference

To apply the pH paper method to ambient aerosols, anotherpotential interference on the captured pH paper color wouldcome from some light-absorbing aerosols such as BC orbrown carbon (BrC). Therefore, we further examined thepotential interference of BC on the predicted pH of lab-generated aerosols. Details regarding the aerosol generationand collection can be found in Sect. 2.2.

Figure 5 shows pHpredict vs. pHreference for the generatedaerosol particles (i.e., (NH4)2SO4−H2SO4−C3H4O4) withand without the co-existence of BC. Note that pHreferencerefers to the pH of bulk solutions used for aerosol generation.

Atmos. Meas. Tech., 13, 6053–6065, 2020 https://doi.org/10.5194/amt-13-6053-2020

G. Li et al.: Determination of aerosol pH by multifactor colorimetric analysis 6061

Figure 5. pH estimation using the type V pH paper for lab-generated aerosols with or without the co-existence of black carbon(BC). pHpredict is calculated with the averaged coefficient vector[a, b, c] derived from five replicate calibration experiments withstandard-buffer-generated aerosol samples. The error bars representthe SD of three replicate experiments. The heights of the orange andblue bars indicate the reported pH ranges measured with pH papersand Raman spectroscopy, respectively, for (NH4)2SO4−H2SO4aerosols with particle sizes in the range of 0.4–2.5 µm in Craig et al.(2018). At pHreference < 2.5, each orange or blue bar has the samepHreference as the orange symbol close to it. Image processing of thecollected aerosol samples follows a similar procedure as describedin Sect. 2.3.

Generally, within the examined pH range no significant dif-ference can be found between the pHpredict of aerosols withBC and that of the aerosols without BC. The linear fitting(i.e., the orange and blue dashed lines in Fig. 5) for each typeof dataset shows that the pHpredict for aerosols with BC isslightly lower than the samples without BC at the low pHside, but an opposite trend can be found in the high pH side.This statistically small difference is further confirmed by run-ning two-sample t tests with Matlab, as shown in Table S4.Even though this difference (≤ 0.5 units) is slight and accept-able, it indicates the existence of potential interferences ofBC on the predicted aerosol pH, and related mechanisms mayneed to be explored in future studies. Note that for our lab ex-periments the adopted BC amount accounted for ∼ 10 % ofthe total aerosol mass, which reflects the typical BC contri-butions in ambient aerosols (Wang et al., 2016b; Chen et al.,2020).

Moreover, both types of aerosols display a lower pHpredictthan pHreference in the low pH range as pHreference< 2.5(Fig. 5). Within the same lower pH range, significantly re-duced aerosol pH (vs. the pH of bulk solutions) predicted byboth pH papers and Raman spectroscopy were also found inCraig et al. (2018) for lab-generated aerosols, as indicatedby the neighbored orange and blue bars in Fig. 5. Their re-sults (Craig et al., 2018) further revealed that the markedlylower pHpredict trend weakened at the higher pH range (i.e.,2.5< pHreference< 4.5; see the orange bars in Fig. 5). The au-

thors argued that the decreased aerosol pH found for smaller-size particles (with aerodynamic diameter < 2.5 µm) couldbe attributed to ammonia partitioning and water loss (Craiget al., 2018). Even with controlled RH for the aerosol dilutionair flow in this study (Sect. 2.2), we cannot totally exclude theimpact of water loss on the predicted aerosol pH, consider-ing that under such a high RH (∼ 90 %) a small differencebetween the RH of the generated aerosol flow and that of thedilution flow may cause non-negligible water exchange be-tween aerosols and the carrying gas.

In addition, the results shown in Fig. 5 further demonstratethe technical feasibility of using our custom-made impactorsfor aerosol collection. More importantly, with this impactorsetup, we could monitor the change in the pH paper color atany sampling time without interrupting the sampling. Thus,when used for future ambient aerosol collection we wouldexpect a small difference between the surrounding environ-ment of aerosols inside the impactors and ambient condi-tions.

4.3 Identification of the needed minimum sampleamount and sampling time for the type V pH paper

The pH of ambient aerosols can be changing due to thevarying atmospheric composition (e.g., some important tracegases like SO2, NO2, NH3, and organic acids) and physi-cal characteristics (e.g., ambient relative humidity, RH, andtemperature, T ). Thus, reflecting the temporal evolution ofaerosol pH requires a pH measurement method with a hightime resolution. As mentioned before, to collect 2 µL of liq-uid aerosol samples, a sampling time of 2–4 h is needed evenusing a high-volume air sampler with a sampling flow rateof several hundred liters per minute. Here, in order to havea higher time resolution and/or a lower sampling flow rate,we further identified the minimum sample volume needed togenerate a measurable color change on the type V pH pa-per. Figure 3b shows the results for 0.1 µL of lab-preparedaerosol sample solutions. Similar to the RGB analysis proce-dure used for the 2 µL samples (e.g., in Fig. 3a), the pHpredictin Fig. 3b is calculated with the averaged coefficient vector[a, b, c] derived from three replicate calibration experimentswith 0.1 µL standard buffers (Fig. S8). Generally, pHpredictagrees well with pHreference, with biases (averaged pHpredictvs. pHreference) within 0.5 units. Note that these experimentswere carried out under laboratory conditions with a rela-tively stable RH of 50 %± 1 % and T of 23± 1 ◦C. To avoidfast water exchange between the lab air and our samples aswell as potential interfering effects (absorption or reaction)caused by the lab air, the 0.1 µL samples were transferred(through a pipette) directly onto the pH paper surface, andeach sample was immediately photographed (≤∼ 3 s) afterit got contact with the pH paper dye. Due to this extremelysmall sample volume, the influence of lab air on sample pHcould become prominent because a significant sample color

https://doi.org/10.5194/amt-13-6053-2020 Atmos. Meas. Tech., 13, 6053–6065, 2020

6062 G. Li et al.: Determination of aerosol pH by multifactor colorimetric analysis

change was frequently observed after the sample was ex-posed to the lab air for &5 s.

This tiny volume corresponds to a sample mass of∼ 180 µg assuming an effective density of 1.8 gcm−3 forambient aerosols (Sarangi et al., 2016; Geller et al., 2006),which is low compared to the needed minimum particulatemasses in Craig et al. (2018), i.e., ∼ 65 µg to ∼ 2.5 mg forPM2.5 or larger particles with pH from 0–2.5 to 2.5–4. Notethat, as we pipetted 0.1 µL (∼ 180 µg) samples on the type VpH paper, this amount cannot be directly used for estimationsof the time needed for ambient aerosol sampling, which alsodepends on how aerosols will be collected on the pH paper.However, this minimum-sample-amount test could provideus with a general estimation on the lower limit of the neededvolume or mass of collected ambient aerosols, which can fur-ther guide us to search for new techniques or instruments foraerosol collection as well as color recognition. As describedin Sect. 2.2, two custom-made impactors were employed tocollect lab-generated aerosols on the type V pH paper. Sincesampling time determined the amount of collected aerosolsand thereby affected the displayed color on pH papers, anoptimal sampling time of 30 min was identified in this studyby examining the established linearity between pHpredict andpHreference (with the Eq. 5) for aerosols generated from stan-dard buffers. With this sampling time, a good linearity withR2> 0.95 was established (Fig. S9). Taking the samplingtime and the mass concentration of lab-generated aerosols(∼ 800 µgm−3, measured under RH= 14 %) into account,we would infer a sampling time of ∼ 4 h will be neededto generate one predicted aerosol pH for ambient aerosolsunder typically polluted conditions (with PM2.5 mass con-centration around 100 µgm−3). Since this time estimation isbased on a sampling flow rate of 28.6 Lmin−1, future sam-plings with a higher time resolution (e.g.,∼ 1 h) can probablybe achieved by adopting a much larger sampling flow rate(e.g., ∼ 120 Lmin−1). Our preliminary tests have indicatedthat the cutoff size, collection efficiency, and pressure dropof the impactors strongly depended on the sampling flow rateand flow direction (i.e., switch between inlets and outlets ofthe impactors). More characterizations on the impactors willbe done in our future work.

These results confirm the feasibility of the type V pH paperas well as our RGB model for pH estimation of the aerosolsample solutions with a volume even down to 0.1 µL and ofthe lab-generated aerosols collected by impactors. The spec-ulated low sampling time (i.e., ∼ 1 h) at high sampling flowrates and the large pH detection range of the type V pH pa-per highlight its potential for future development of real-timeaerosol pH measurements. Moreover, instead of using a colorchart to calibrate pH papers for each sample (Craig et al.,2018), our results demonstrate that the in situ calibrationmethod of using standard buffers as well as standard-buffer-generated aerosols (independent of different samples) canderive an averaged coefficient vector [a, b, c] which can beuniformly applied to pH prediction of different samples pro-

vided the photographing conditions are kept constant. Thisunique feature further facilitates the application of the type VpH paper under ambient cases.

5 Conclusions

We proposed a new model to establish the correlation be-tween the color of droplet or aerosol samples on pH-indicatorpapers and their measured pH. The model was based onRGB analysis of the images of samples. Comparison of ourmodel and another two RGB models verified the high re-liability of our model. Using our RGB model, good agree-ment between the model-predicted pH (pHpredict) and refer-ence pH (pHreference) for pH paper color charts as well asstandard buffers was observed for all the tested types of pHpapers. Different types of pH papers with larger pH detec-tion ranges were further examined for their performance topredict the pH of 2 µL droplet samples with varying inor-ganic or organic compositions common in ambient aerosols.Only the type V pH paper (with a pH range of ∼ 0–6) per-formed well and therefore was further used to estimate thepH of lab-generated aerosols. The results showed that, evenunder the potential interference of BC, the type V pH pa-per could derive aerosol pH with an uncertainty within 0.5units, suggesting that it deserves practical applications forpH measurements of ambient aerosols. The minimum liquidsample mass or volume needed for the type V pH paper wasidentified as ∼ 180 µg or 0.1 µL. And the current-stage testson aerosol collection and pH estimation under lab conditionshelped to infer that an ambient sampling time of ∼ 4 h willbe needed for typically polluted conditions, which, however,could probably be further improved to ∼ 1 h by using a highsampling flow rate, whereas the other pH paper types mightsuffer from some chemical interferences during pH measure-ments and therefore could generate large biases for the mea-sured pH of aerosols. The routine procedure of using pH pa-pers to estimate a sample pH was also examined in a casestudy using the type IV pH paper. The results showed thatreferring to the color chart for pH estimation (i.e., the color-chart-calibration method) might cause a bias in the predictedpH. To use the pH papers in a more proper and accurateway, here we further demonstrated that the in situ calibra-tion method of using standard buffers and standard-buffer-generated aerosols (independent of different samples) couldderive an averaged coefficient vector [a, b, c], which can beuniformly applied to pH prediction of different droplet andaerosol samples provided the photographing conditions arekept constant.

Code and data availability. The underlying research data andMatlab code can be accessed upon contact with Guo Li(guo.li@mpic.de), Yafang Cheng (yafang.cheng@mpic.de), orHang Su (h.su@mpic.de).

Atmos. Meas. Tech., 13, 6053–6065, 2020 https://doi.org/10.5194/amt-13-6053-2020

G. Li et al.: Determination of aerosol pH by multifactor colorimetric analysis 6063

Supplement. The supplement related to this article is available on-line at: https://doi.org/10.5194/amt-13-6053-2020-supplement.

Author contributions. YC and HS conceived and led the study. GLperformed experiments and data analysis. YC, HS, UP, UK, NM,GZ, ML, and TK discussed the results. GL and YC wrote the paperwith inputs from all co-authors.

Competing interests. The authors declare that they have no conflictof interest.

Acknowledgements. This study was supported by the Max PlanckSociety (MPG). Guo Li would like to thank Ping Zhang, JinqianZhai, and Yunkun Lang for their very helpful discussions concern-ing the RGB model development.

Financial support. This research has been supported by the Na-tional Natural Science Foundation of China (grant no. 91644218),the National Key Research and Development Program of China(grant no. 2017YFC0210104), and the China Scholarship Council(CSC).

The article processing charges for this open-accesspublication were covered by the Max Planck Society.

Review statement. This paper was edited by Mingjin Tang and re-viewed by Rodney Weber and one anonymous referee.

References

Abbatt, J. P. D., Broekhuizen, K., and Pradeep Kumar, P.:Cloud condensation nucleus activity of internally mixed am-monium sulfate/organic acid aerosol particles, Atmos. Environ.,39, 4767–4778, https://doi.org/10.1016/j.atmosenv.2005.04.029,2005.

Ahrens, L., Harner, T., Shoeib, M., Lane, D. A., and Murphy, J.G.: Improved Characterization of Gas–Particle Partitioning forPer- and Polyfluoroalkyl Substances in the Atmosphere UsingAnnular Diffusion Denuder Samplers, Environ. Sci. Technol., 46,7199–7206, https://doi.org/10.1021/es300898s, 2012.

Baltensperger, U., Dommen, J., Alfarra, M. R., Duplissy, J.,Gaeggeler, K., Metzger, A., Facchini, M. C., Decesari, S., Fi-nessi, E., Reinnig, C., Schott, M., Warnke, J., Hoffmann, T.,Klatzer, B., Puxbaum, H., Geiser, M., Savi, M., Lang, D.,Kalberer, M., and Geiser, T.: Combined Determination of theChemical Composition and of Health Effects of Secondary Or-ganic Aerosols: The POLYSOA Project, J. Aerosol Med. Pulm.D., 21, 145–154, https://doi.org/10.1089/jamp.2007.0655, 2008.

Cai, J., Zhi, G., Yu, Z., Nie, P., Gligorovski, S., Zhang, Y., Zhu, L.,Guo, X., Li, P., He, T., He, Y., Sun, J., and Zhang, Y.: Spec-tral changes induced by pH variation of aqueous extracts de-rived from biomass burning aerosols: Under dark and in pres-

ence of simulated sunlight irradiation, Atmos. Environ., 185, 1–6, https://doi.org/10.1016/j.atmosenv.2018.04.037, 2018.

Chen, L., Zhang, F., Yan, P., Wang, X., Sun, L., Li, Y., Zhang,X., Sun, Y., and Li, Z.: The large proportion of black carbon(BC)-containing aerosols in the urban atmosphere, Environ. Pol-lut., 263, 114507, https://doi.org/10.1016/j.envpol.2020.114507,2020.

Cheng, Y., Zheng, G., Wei, C., Mu, Q., Zheng, B., Wang, Z., Gao,M., Zhang, Q., He, K., Carmichael, G., Pöschl, U., and Su,H.: Reactive nitrogen chemistry in aerosol water as a sourceof sulfate during haze events in China, Science Advances, 2,e1601530, https://doi.org/10.1126/sciadv.1601530, 2016.

Chowdhury, P. H., He, Q., Lasitza Male, T., Brune, W. H.,Rudich, Y., and Pardo, M.: Exposure of Lung Epithelial Cellsto Photochemically Aged Secondary Organic Aerosol ShowsIncreased Toxic Effects, Environ. Sci. Tech. Let., 5, 424–430,https://doi.org/10.1021/acs.estlett.8b00256, 2018.

Clegg, S. L. and Seinfeld, J. H.: Thermodynamic Models ofAqueous Solutions Containing Inorganic Electrolytes and Di-carboxylic Acids at 298.15 K. 2. Systems Including Dis-sociation Equilibria, J. Phys. Chem. A, 110, 5718–5734,https://doi.org/10.1021/jp056150j, 2006a.

Clegg, S. L. and Seinfeld, J. H.: Thermodynamic Models ofAqueous Solutions Containing Inorganic Electrolytes and Di-carboxylic Acids at 298.15 K. 1. The Acids as Nondisso-ciating Components, J. Phys. Chem. A, 110, 5692–5717,https://doi.org/10.1021/jp056149k, 2006b.

Craig, R. L., Nandy, L., Axson, J. L., Dutcher, C. S., andAult, A. P.: Spectroscopic Determination of Aerosol pHfrom Acid–Base Equilibria in Inorganic, Organic, andMixed Systems, J. Phys. Chem. A, 121, 5690–5699,https://doi.org/10.1021/acs.jpca.7b05261, 2017.

Craig, R. L., Peterson, P. K., Nandy, L., Lei, Z., Hossain, M.A., Camarena, S., Dodson, R. A., Cook, R. D., Dutcher, C.S., and Ault, A. P.: Direct Determination of Aerosol pH:Size-Resolved Measurements of Submicrometer and Supermi-crometer Aqueous Particles, Anal. Chem., 90, 11232–11239,https://doi.org/10.1021/acs.analchem.8b00586, 2018.

Dallemagne, M. A., Huang, X. Y., and Eddingsaas, N. C.: Varia-tion in pH of Model Secondary Organic Aerosol during Liquid–Liquid Phase Separation, J. Phys. Chem. A, 120, 2868–2876,https://doi.org/10.1021/acs.jpca.6b00275, 2016.

Dinar, E., Anttila, T., and Rudich, Y.: CCN Activity and Hy-groscopic Growth of Organic Aerosols Following ReactiveUptake of Ammonia, Environ. Sci. Technol., 42, 793–799,https://doi.org/10.1021/es071874p, 2008.

Falkovich, A. H., Graber, E. R., Schkolnik, G., Rudich, Y., Maen-haut, W., and Artaxo, P.: Low molecular weight organic acidsin aerosol particles from Rondônia, Brazil, during the biomass-burning, transition and wet periods, Atmos. Chem. Phys., 5, 781–797, https://doi.org/10.5194/acp-5-781-2005, 2005.

Fang, T., Guo, H., Zeng, L., Verma, V., Nenes, A., and We-ber, R. J.: Highly Acidic Ambient Particles, Soluble Met-als, and Oxidative Potential: A Link between Sulfate andAerosol Toxicity, Environ. Sci. Technol., 51, 2611–2620,https://doi.org/10.1021/acs.est.6b06151, 2017.

Fillion, N., Probst, A., and Probst, J.-L.: Dissolved organic mattercontribution to rain water, throughfall and soil solution chem-istry, Analusis, 27, 409–413, 1999.

https://doi.org/10.5194/amt-13-6053-2020 Atmos. Meas. Tech., 13, 6053–6065, 2020

6064 G. Li et al.: Determination of aerosol pH by multifactor colorimetric analysis

Fountoukis, C. and Nenes, A.: ISORROPIA II: a computa-tionally efficient thermodynamic equilibrium model for K+–Ca2+–Mg2+–NH+4 –Na+–SO2−

4 –NO−3 –Cl−–H2O aerosols, At-mos. Chem. Phys., 7, 4639–4659, https://doi.org/10.5194/acp-7-4639-2007, 2007.

Gao, S., Ng, N. L., Keywood, M., Varutbangkul, V., Bahreini,R., Nenes, A., He, J., Yoo, K. Y., Beauchamp, J. L.,Hodyss, R. P., Flagan, R. C., and Seinfeld, J. H.: Par-ticle Phase Acidity and Oligomer Formation in SecondaryOrganic Aerosol, Environ. Sci. Technol., 38, 6582–6589,https://doi.org/10.1021/es049125k, 2004.

Gaston, C. J., Riedel, T. P., Zhang, Z., Gold, A., Surratt, J. D., andThornton, J. A.: Reactive Uptake of an Isoprene-Derived Epoxy-diol to Submicron Aerosol Particles, Environ. Sci. Technol., 48,11178–11186, https://doi.org/10.1021/es5034266, 2014.

Geller, M., Biswas, S., and Sioutas, C.: Determination ofParticle Effective Density in Urban Environments witha Differential Mobility Analyzer and Aerosol Parti-cle Mass Analyzer, Aerosol Sci. Tech., 40, 709–723,https://doi.org/10.1080/02786820600803925, 2006.

Han, Y., Stroud, C. A., Liggio, J., and Li, S.-M.: The ef-fect of particle acidity on secondary organic aerosol for-mation from α-pinene photooxidation under atmosphericallyrelevant conditions, Atmos. Chem. Phys., 16, 13929–13944,https://doi.org/10.5194/acp-16-13929-2016, 2016.

Hennigan, C. J., Izumi, J., Sullivan, A. P., Weber, R. J., and Nenes,A.: A critical evaluation of proxy methods used to estimate theacidity of atmospheric particles, Atmos. Chem. Phys., 15, 2775–2790, https://doi.org/10.5194/acp-15-2775-2015, 2015.

Hinrichs, R. Z., Buczek, P., and Trivedi, J. J.: Solar Absorp-tion by Aerosol-Bound Nitrophenols Compared to Aqueous andGaseous Nitrophenols, Environ. Sci. Technol., 50, 5661–5667,https://doi.org/10.1021/acs.est.6b00302, 2016.

Hourant, P.: General Properties of the Alkaline Phosphates: – MajorFood and Technical Applications, Phosphorus Research Bulletin,15, 85–94, https://doi.org/10.3363/prb1992.15.0_85, 2004.

Iinuma, Y., Böge, O., Gnauk, T., and Herrmann, H.: Aerosol-chamber study of the α-pinene/O3 reaction: influence of parti-cle acidity on aerosol yields and products, Atmos. Environ., 38,761–773, https://doi.org/10.1016/j.atmosenv.2003.10.015, 2004.

Jang, M., Cao, G., and Paul, J.: Colorimetric Particle Acid-ity Analysis of Secondary Organic Aerosol Coating on Sub-micron Acidic Aerosols, Aerosol Sci. Tech., 42, 409–420,https://doi.org/10.1080/02786820802154861, 2008.

Jimenez, J. L., Canagaratna, M. R., Donahue, N. M., Prevot, A. S.H., Zhang, Q., Kroll, J. H., DeCarlo, P. F., Allan, J. D., Coe,H., Ng, N. L., Aiken, A. C., Docherty, K. S., Ulbrich, I. M.,Grieshop, A. P., Robinson, A. L., Duplissy, J., Smith, J. D., Wil-son, K. R., Lanz, V. A., Hueglin, C., Sun, Y. L., Tian, J., Laak-sonen, A., Raatikainen, T., Rautiainen, J., Vaattovaara, P., Ehn,M., Kulmala, M., Tomlinson, J. M., Collins, D. R., Cubison, M.J., Dunlea, J., Huffman, J. A., Onasch, T. B., Alfarra, M. R.,Williams, P. I., Bower, K., Kondo, Y., Schneider, J., Drewnick,F., Borrmann, S., Weimer, S., Demerjian, K., Salcedo, D., Cot-trell, L., Griffin, R., Takami, A., Miyoshi, T., Hatakeyama, S.,Shimono, A., Sun, J. Y., Zhang, Y. M., Dzepina, K., Kimmel,J. R., Sueper, D., Jayne, J. T., Herndon, S. C., Trimborn, A.M., Williams, L. R., Wood, E. C., Middlebrook, A. M., Kolb,C. E., Baltensperger, U., and Worsnop, D. R.: Evolution of Or-

ganic Aerosols in the Atmosphere, Science, 326, 1525–1529,https://doi.org/10.1126/science.1180353, 2009.

Keene, W. C., Pszenny, A. A. P., Maben, J. R., and Sander, R.: Vari-ation of marine aerosol acidity with particle size, Geophys. Res.Lett., 29, 5-1–5-4, https://doi.org/10.1029/2001gl013881, 2002.

Keene, W. C., Pszenny, A. A. P., Maben, J. R., Stevenson, E., andWall, A.: Closure evaluation of size-resolved aerosol pH in theNew England coastal atmosphere during summer, J. Geophys.Res., 109, D23307, https://doi.org/10.1029/2004JD004801,2004.

Kondo, Y., Sahu, L., Moteki, N., Khan, F., Takegawa, N., Liu,X., Koike, M., and Miyakawa, T.: Consistency and Traceabil-ity of Black Carbon Measurements Made by Laser-Induced In-candescence, Thermal-Optical Transmittance, and Filter-BasedPhoto-Absorption Techniques, Aerosol Sci. Tech., 45, 295–312,https://doi.org/10.1080/02786826.2010.533215, 2011.

Koutrakis, P., Wolfson, J. M., and Spengler, J. D.: An improvedmethod for measuring aerosol strong acidity: Results from anine-month study in St Louis, Missouri and Kingston, Tennessee,Atmos. Environ., 22, 157–162, https://doi.org/10.1016/0004-6981(88)90308-3, 1988.

Li, H., Liu, F., Kang, L., and Zheng, M.: Study on thebuffering capacity of wort, J. I. Brewing, 122, 138–142,https://doi.org/10.1002/jib.286, 2016.

Li, J. and Jang, M.: Aerosol Acidity Measurement Us-ing Colorimetry Coupled With a Reflectance UV-Visible Spectrometer, Aerosol Sci. Tech., 46, 833–842,https://doi.org/10.1080/02786826.2012.669873, 2012.

Losey, D. J., Ott, E.-J. E., and Freedman, M. A.: Ef-fects of High Acidity on Phase Transitions of an Or-ganic Aerosol, J. Phys. Chem. A, 122, 3819–3828,https://doi.org/10.1021/acs.jpca.8b00399, 2018.

McNeill, V. F.: Atmospheric Aerosols: Clouds, Chemistry,and Climate, Annu. Rev. Chem. Biomol., 8, 427–444,https://doi.org/10.1146/annurev-chembioeng-060816-101538,2017.

Northcross, A. L. and Jang, M.: Heterogeneous SOAyield from ozonolysis of monoterpenes in the pres-ence of inorganic acid, Atmos. Environ., 41, 1483–1493,https://doi.org/10.1016/j.atmosenv.2006.10.009, 2007.

Padró, J. M., Acquaviva, A., Tascon, M., Gagliardi, L. G.,and Castells, C. B.: Effect of temperature and solvent com-position on acid dissociation equilibria, I: Sequenced pssKadetermination of compounds commonly used as buffers inhigh performance liquid chromatography coupled to massspectroscopy detection, Anal. Chim. Acta, 725, 87–94,https://doi.org/10.1016/j.aca.2012.03.015, 2012.

Pöschl, U.: Atmospheric Aerosols: Composition, Transformation,Climate and Health Effects, Angew. Chem. Int. Edit., 44, 7520–7540, https://doi.org/10.1002/anie.200501122, 2005.

Pósfai, M. and Buseck, P. R.: Nature and Climate Effects of Individ-ual Tropospheric Aerosol Particles, Annu. Rev. Earth Pl. Sc., 38,17–43, https://doi.org/10.1146/annurev.earth.031208.100032,2010.

Pszenny, A. A. P., Moldanová, J., Keene, W. C., Sander, R.,Maben, J. R., Martinez, M., Crutzen, P. J., Perner, D., andPrinn, R. G.: Halogen cycling and aerosol pH in the Hawai-ian marine boundary layer, Atmos. Chem. Phys., 4, 147–168,https://doi.org/10.5194/acp-4-147-2004, 2004.

Atmos. Meas. Tech., 13, 6053–6065, 2020 https://doi.org/10.5194/amt-13-6053-2020

G. Li et al.: Determination of aerosol pH by multifactor colorimetric analysis 6065

Rindelaub, J. D., Craig, R. L., Nandy, L., Bondy, A. L., Dutcher, C.S., Shepson, P. B., and Ault, A. P.: Direct Measurement of pH inIndividual Particles via Raman Microspectroscopy and Variationin Acidity with Relative Humidity, J. Phys. Chem. A, 120, 911–917, https://doi.org/10.1021/acs.jpca.5b12699, 2016.

Sarangi, B., Aggarwal, S. G., Sinha, D., and Gupta, P. K.: Aerosoleffective density measurement using scanning mobility par-ticle sizer and quartz crystal microbalance with the estima-tion of involved uncertainty, Atmos. Meas. Tech., 9, 859–875,https://doi.org/10.5194/amt-9-859-2016, 2016.

Selva Kumar, R., Kumar, S. K. A., Vijayakrishna, K., Sivaramakr-ishna, A., Brahmmananda Rao, C. V. S., Sivaraman, N., and Sa-hoo, S. K.: Development of the Smartphone-Assisted Colorimet-ric Detection of Thorium by Using New Schiff’s Base and ItsApplications to Real Time Samples, Inorg. Chem., 57, 15270–15279, https://doi.org/10.1021/acs.inorgchem.8b02564, 2018.

Shi, G., Xu, J., Shi, X., Liu, B., Bi, X., Xiao, Z., Chen, K.,Wen, J., Dong, S., Tian, Y., Feng, Y., Yu, H., Song, S., Zhao,Q., Gao, J., and Russell, A. G.: Aerosol pH dynamics duringintense haze periods in an urban environment in China: useof detailed, hourly, speciated observations to study the role ofammonia availability and secondary aerosol formation and ur-ban environment, J. Geophys. Res.-Atmos., 124, 9730–9742,https://doi.org/10.1029/2018JD029976, 2019.

Shiraiwa, M., Ueda, K., Pozzer, A., Lammel, G., Kampf, C. J.,Fushimi, A., Enami, S., Arangio, A. M., Fröhlich-Nowoisky, J.,Fujitani, Y., Furuyama, A., Lakey, P. S. J., Lelieveld, J., Lucas,K., Morino, Y., Pöschl, U., Takahama, S., Takami, A., Tong, H.,Weber, B., Yoshino, A., and Sato, K.: Aerosol Health Effectsfrom Molecular to Global Scales, Environ. Sci. Technol., 51,13545–13567, https://doi.org/10.1021/acs.est.7b04417, 2017.

Song, S., Gao, M., Xu, W., Shao, J., Shi, G., Wang, S., Wang,Y., Sun, Y., and McElroy, M. B.: Fine-particle pH for Bei-jing winter haze as inferred from different thermodynamicequilibrium models, Atmos. Chem. Phys., 18, 7423–7438,https://doi.org/10.5194/acp-18-7423-2018, 2018.

Su, C.-H., Fu, C.-C., Chang, Y.-C., Nair, G. R., Ye, J.-L., Chu, I.-M., and Wu, W.-T.: Simultaneous estimation of chlorophyll aand lipid contents in microalgae by three-color analysis, Biotech-nol. Bioeng., 99, 1034–1039, https://doi.org/10.1002/bit.21623,2008.

Su, H., Cheng, Y., and Pöschl, U.: New Multiphase Chemical Pro-cesses Influencing Atmospheric Aerosols, Air Quality, and Cli-mate in the Anthropocene, Accounts Chem. Res., 53, 2034–2043, https://doi.org/10.1021/acs.accounts.0c00246, 2020.

Surratt, J. D., Lewandowski, M., Offenberg, J. H., Jaoui,M., Kleindienst, T. E., Edney, E. O., and Seinfeld, J. H.:Effect of Acidity on Secondary Organic Aerosol Forma-tion from Isoprene, Environ. Sci. Technol., 41, 5363–5369,https://doi.org/10.1021/es0704176, 2007.

Surratt, J. D., Gómez-González, Y., Chan, A. W. H., Vermeylen, R.,Shahgholi, M., Kleindienst, T. E., Edney, E. O., Offenberg, J. H.,Lewandowski, M., Jaoui, M., Maenhaut, W., Claeys, M., Flagan,R. C., and Seinfeld, J. H.: Organosulfate Formation in BiogenicSecondary Organic Aerosol, J. Phys. Chem. A, 112, 8345–8378,https://doi.org/10.1021/jp802310p, 2008.

Tao, W., Su, H., Zheng, G., Wang, J., Wei, C., Liu, L., Ma, N., Li,M., Zhang, Q., Pöschl, U., and Cheng, Y.: Aerosol pH and chem-ical regimes of sulfate formation in aerosol water during winter

haze in the North China Plain, Atmos. Chem. Phys., 20, 11729–11746, https://doi.org/10.5194/acp-20-11729-2020, 2020.

von Glasow, R. and Sander, R.: Variation of sea salt aerosolpH with relative humidity, Geophys. Res. Lett., 28, 247–250,https://doi.org/10.1029/2000GL012387, 2001.

von Schneidemesser, E., Monks, P. S., Allan, J. D., Bruhwiler, L.,Forster, P., Fowler, D., Lauer, A., Morgan, W. T., Paasonen, P.,Righi, M., Sindelarova, K., and Sutton, M. A.: Chemistry and theLinkages between Air Quality and Climate Change, Chem. Rev.,115, 3856–3897, https://doi.org/10.1021/acs.chemrev.5b00089,2015.

Wan, Z., Zhong, L., Pan, Y., Li, H., Zou, Q., Su, K., and Wang,P.: Portable Microplate Analyzer with a Thermostatic ChamberBased on a Smartphone for On-site Rapid Detection, Anal. Sci.,33, 1291–1296, https://doi.org/10.2116/analsci.33.1291, 2017.

Wang, G., Zhang, R., Gomez, M. E., Yang, L., Levy Zamora, M.,Hu, M., Lin, Y., Peng, J., Guo, S., Meng, J., Li, J., Cheng, C.,Hu, T., Ren, Y., Wang, Y., Gao, J., Cao, J., An, Z., Zhou, W.,Li, G., Wang, J., Tian, P., Marrero-Ortiz, W., Secrest, J., Du, Z.,Zheng, J., Shang, D., Zeng, L., Shao, M., Wang, W., Huang, Y.,Wang, Y., Zhu, Y., Li, Y., Hu, J., Pan, B., Cai, L., Cheng, Y., Ji,Y., Zhang, F., Rosenfeld, D., Liss, P. S., Duce, R. A., Kolb, C. E.,and Molina, M. J.: Persistent sulfate formation from London Fogto Chinese haze, P. Natl. Acad. Sci. USA, 113, 13630–13635,https://doi.org/10.1073/pnas.1616540113, 2016a.

Wang, Q., Huang, R.-J., Zhao, Z., Cao, J., Ni, H., Tie, X., Zhao,S., Su, X., Han, Y., Shen, Z., Wang, Y., Zhang, N., Zhou,Y., and Corbin, J. C.: Physicochemical characteristics of blackcarbon aerosol and its radiative impact in a polluted urbanarea of China, J. Geophys. Res.-Atmos., 121, 12505–12519,https://doi.org/10.1002/2016JD024748, 2016b.

Wei, H., Vejerano, E. P., Leng, W., Huang, Q., Willner, M. R.,Marr, L. C., and Vikesland, P. J.: Aerosol microdroplets exhibita stable pH gradient, P. Natl. Acad. Sci. USA, 115, 7272–7277,https://doi.org/10.1073/pnas.1720488115, 2018.

Yadav, S. P., Ibaraki, Y., and Dutta Gupta, S.: Estimation of thechlorophyll content of micropropagated potato plants using RGBbased image analysis, Plant Cell Tiss. Org., 100, 183–188,https://doi.org/10.1007/s11240-009-9635-6, 2010.

Zhang, F., Wang, Y., Peng, J., Chen, L., Sun, Y., Duan, L., Ge, X.,Li, Y., Zhao, J., Liu, C., Zhang, X., Zhang, G., Pan, Y., Wang, Y.,Zhang, A. L., Ji, Y., Wang, G., Hu, M., Molina, M. J., and Zhang,R.: An unexpected catalyst dominates formation and radiativeforcing of regional haze, P. Natl. Acad. Sci. USA, 117, 3960–3966, https://doi.org/10.1073/pnas.1919343117, 2020.

Zheng, G., Su, H., Wang, S., Andreae, M. O., Pöschl, U.,and Cheng, Y.: Multiphase buffer theory explains contrastsin atmospheric aerosol acidity, Science, 369, 1374–1377,https://doi.org/10.1126/science.aba3719, 2020.

Zheng, G. J., Duan, F. K., Su, H., Ma, Y. L., Cheng, Y., Zheng,B., Zhang, Q., Huang, T., Kimoto, T., Chang, D., Pöschl, U.,Cheng, Y. F., and He, K. B.: Exploring the severe winter haze inBeijing: the impact of synoptic weather, regional transport andheterogeneous reactions, Atmos. Chem. Phys., 15, 2969–2983,https://doi.org/10.5194/acp-15-2969-2015, 2015.

https://doi.org/10.5194/amt-13-6053-2020 Atmos. Meas. Tech., 13, 6053–6065, 2020